|Faculty or Centre||Faculté des Sciences, de la Technologie et de la Communication|
Weicker Building, Université du Luxembourg
4, rue Alphonse Weicker
|Telephone||(+352) 46 66 44 5573|
|Fax||(+352) 46 66 44 35573|
Please visit my personal web page: http://www.darkrsw.net
As the software is getting complex, programmers suffer from a huge number of bugs everyday. Automated program repair focuses on how to generate program patches so that developers can pay more attention to other important tasks. My approach looks up human-written patches and extracts fix patterns from them. These patterns guide how to generate patches automatically. The benefit of this approach is that it can generate more realistic patches than existing techniques.
Because software crashes is one of the catastrophic failure, programmers must fix these defects as highest priority. However, the huge number of crashing bugs is submitted and the developers are already outnumbered by the bugs. To handle this problem, my approach prioritizes "top crashes", which will happen more frequently, at an early stage. This approach uses history, complexity, and social network analysis metrics to predict top crashes. We applied this approach to Mozilla applications and the result showed that it can predict top crashes with 68~80% accuracy.
Do we need to write down every program we need? Many programmers already wrote a large number of programs. We can reuse them! However, reusing program code is another tedious task. Recently I am focusing on automatically generate programs from existing programs. This approach can reduce a huge amount of programming effort and make programmers focus on defining <it>what the program has to do rather than how to do it.
Last updated on: Wednesday, 03 February 2016
in Empirical Software Engineering (2016), 21(2), 565-604
; ; ; ; ;
; ; ; ; ;
Scientific Conference (2014, December 01)